A geological disaster warning method and system based on big data analysis

Abstract

The present invention discloses a geological disaster early warning method and system based on big data analysis, specifically relating to the technical field of geological disaster early warning; by using satellite remote sensing monitoring to obtain the time series data of geological deformation in the target area, and adopting the random forest algorithm to evaluate the feature importance and divide the regional geological risk level; at the same time, performing wavelet transform and long short-term memory network modeling on the rainfall spatio-temporal data of meteorological monitoring stations to evaluate the potential impact of rainfall; screening high-risk areas, constructing an association model between the change rate of groundwater level and the strength of rock and soil mass, and adopting the support vector machine algorithm to evaluate the landslide probability; finally, using the analytic hierarchy process to analyze the results of geological risk level division, the potential impact degree of rainfall on the target area and the landslide occurrence probability, calculating the comprehensive risk index, and comparing it with the comprehensive risk threshold to determine whether to trigger a hierarchical early warning, realizing data fusion and intelligent analysis, early warning, and providing effective help for disaster prevention and mitigation.

Description

Technical Field

[0001] The present invention relates to the technical field of geological disaster early warning, and more specifically, the present invention relates to a geological disaster early warning method and system based on big data analysis. Background Art

[0002] With the rapid development of the economic society and the acceleration of the urbanization process, the regional geological environment is becoming increasingly complex, and geological disasters occur frequently. Traditional methods have deficiencies in aspects such as multi-source data fusion, dynamic risk assessment, and hierarchical early warning, and cannot comprehensively reflect the complexity of the geological environment and the linkage effects of external factors such as rainfall. There are problems such as information islands, data lags, and low model accuracy, making it difficult to achieve early and accurate prediction of disasters.

[0003] To solve the above problems, a technical solution is provided now. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a geological disaster early warning method and system based on big data analysis to solve the problems raised in the above background art.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A geological disaster early warning method based on big data analysis, comprising the following steps:

[0007] Obtain the time series data of geological deformation in the target area through satellite remote sensing monitoring, and use the random forest algorithm to evaluate the feature importance of the geological deformation data, and divide the geological risk levels of the target area;

[0008] Perform wavelet transform analysis on the rainfall spatio-temporal distribution data of the meteorological monitoring station, establish a rainfall trend prediction model based on the long short-term memory neural network, and evaluate the potential impact degree of rainfall on the target area;

[0009] Based on the geological risk level division result and the potential impact degree of rainfall on the target area, screen out high-risk trigger areas;

[0010] For the high-risk trigger areas, by constructing an association model between the groundwater level change rate and the geotechnical strength parameters, use the support vector machine algorithm to evaluate the probability of landslide occurrence;

[0011] Use the analytic hierarchy process to analyze the geological risk level division result, the potential impact degree of rainfall on the target area, and the probability of landslide occurrence, evaluate the comprehensive risk degree of the target area, and determine whether to trigger a hierarchical early warning signal.

[0012] In a preferred embodiment, time-series data of geological deformation in the target area is obtained through satellite remote sensing monitoring, and the random forest algorithm is used to evaluate the feature importance of the geological deformation data to divide the geological risk levels of the target area, specifically as follows:

[0013] Obtain multi-source satellite images and extract the original observation data;

[0014] Denoise and interpolate the original observation data;

[0015] Select geological attributes and historical disaster situations and uniformly label the features;

[0016] Use the random forest to evaluate the feature importance and screen the key factors;

[0017] Use the convolutional neural network to identify the reflection pattern and interference features;

[0018] Based on the screened factors, divide the geological risk levels of the target area.

[0019] In a preferred embodiment, wavelet transform analysis is performed on the rainfall spatio-temporal distribution data of the meteorological monitoring stations, and a rainfall trend prediction model is established based on the long short-term memory neural network to evaluate the potential impact degree of rainfall on the target area, specifically as follows:

[0020] Collect and integrate the rainfall spatio-temporal distribution data of the meteorological monitoring stations in the target area;

[0021] Use wavelet transform to perform multi-scale time-frequency decomposition on the rainfall data;

[0022] Construct a long short-term memory network model, train and verify the rainfall trend prediction samples;

[0023] Calculate the regional rainfall impact coefficient according to the prediction results;

[0024] Based on the rainfall impact coefficient, evaluate the potential impact degree of rainfall on the target area.

[0025] In a preferred embodiment, based on the rainfall impact coefficient, evaluate the potential impact degree of rainfall on the target area, specifically as follows:

[0026] Preset a rainfall impact coefficient threshold and compare the rainfall impact coefficient with the rainfall impact coefficient threshold:

[0027] When the rainfall impact coefficient is greater than the rainfall impact coefficient threshold, it indicates that the potential impact degree of the rainfall event on the target area is high;

[0028] When the rainfall impact coefficient is less than or equal to the rainfall impact coefficient threshold, it indicates that the potential impact degree of the rainfall event on the target area is low.

[0029] In a preferred embodiment, based on the geological risk level division result and the potential impact degree of rainfall on the target area, high-risk trigger areas are screened, specifically as follows:

[0030] When the geological risk level of the target area is level three, the target area is determined as a high-risk trigger area; when the geological risk level of the target area is level two and the rainfall impact coefficient is greater than the rainfall impact coefficient threshold, the target area is determined as a high-risk trigger area; otherwise, the target area is determined as a non-high-risk trigger area.

[0031] In a preferred embodiment, for high-risk trigger areas, by constructing an association model between the groundwater level change rate and the geotechnical strength parameters, the support vector machine algorithm is used to evaluate the landslide occurrence probability, specifically as follows:

[0032] Collect monitoring data on the groundwater level change rate;

[0033] Select geotechnical strength elements and construct association factors;

[0034] Establish an association model between the groundwater level and geotechnical strength;

[0035] Train the support vector machine using the kernel function optimization method;

[0036] Input the association model parameters to evaluate the landslide occurrence probability.

[0037] In a preferred embodiment, the analytic hierarchy process is used to analyze the geological risk level division result, the potential impact degree of rainfall on the target area, and the landslide occurrence probability, evaluate the comprehensive risk degree of the target area, and determine whether to trigger a graded warning signal, specifically as follows:

[0038] Preset a comprehensive risk threshold and compare the comprehensive risk index with the comprehensive risk threshold:

[0039] When the comprehensive risk index is greater than or equal to the comprehensive risk threshold, it indicates that the comprehensive risk degree of the target area is high and a graded warning signal needs to be triggered;

[0040] When the comprehensive risk index is less than the comprehensive risk threshold, it indicates that the comprehensive risk degree of the target area is low and a graded warning signal does not need to be triggered.

[0041] On the other hand, the present invention provides a geological disaster warning system based on big data analysis, including a risk level division module, an impact degree evaluation module, a regional screening module, a landslide probability evaluation module, and a warning signal judgment module;

[0042] The risk level division module obtains the geological deformation time series data of the target area through satellite remote sensing monitoring, and uses the random forest algorithm to evaluate the feature importance of the geological deformation data and divide the geological risk level of the target area;

[0043] The impact degree evaluation module performs wavelet transform analysis on the rainfall spatio-temporal distribution data of meteorological monitoring stations, establishes a rainfall trend prediction model based on long short-term memory neural network, and evaluates the potential impact degree of rainfall on the target area;

[0044] The area screening module screens high-risk trigger areas based on the geological risk level division results and the potential impact degree of rainfall on the target area;

[0045] For high-risk trigger areas, the landslide probability evaluation module evaluates the landslide occurrence probability by constructing a correlation model between the groundwater level change rate and the geotechnical strength parameters and using the support vector machine algorithm;

[0046] The early warning signal judgment module uses the analytic hierarchy process to analyze the geological risk level division results, the potential impact degree of rainfall on the target area, and the landslide occurrence probability, evaluates the comprehensive risk degree of the target area, and judges whether to trigger a hierarchical early warning signal.

[0047] The technical effects and advantages of a geological disaster early warning method and system based on big data analysis according to the present invention:

[0048] By adopting big data fusion technology, comprehensively utilizing satellite remote sensing, meteorological monitoring and sensor data, multi-level and dynamic evaluation of geological disaster risks is realized. Through models such as random forest, long short-term memory neural network and support vector machine, geological deformation, rainfall trend and landslide probability are accurately quantified, improving the early warning accuracy and response speed. Using the analytic hierarchy process to comprehensively analyze each risk index can judge in real time whether to trigger a hierarchical early warning signal, effectively overcoming the problems of early warning lag and data isolation of traditional methods. It shows excellent robustness and adaptability under complex geological environments and extreme climate conditions, provides a scientific basis and decision-making support for disaster prevention and mitigation, and at the same time promotes the intelligent and informatized process of geological disaster early warning technology, having broad application prospects and social and economic benefits. Brief Description of the Drawings

[0049] Figure 1 It is a schematic diagram of a geological disaster early warning method based on big data analysis according to the present invention;

[0050] Figure 2 It is a schematic diagram of the structure of a geological disaster early warning system based on big data analysis according to the present invention. Detailed Embodiments

[0051] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0052] Embodiment 1:

[0053] Figure 1 A geological disaster early warning method based on big data analysis of the present invention is provided, which includes the following steps:

[0054] Obtain the time series data of geological deformation in the target area through satellite remote sensing monitoring, and use the random forest algorithm to evaluate the feature importance of the geological deformation data to divide the geological risk level of the target area;

[0055] Perform wavelet transform analysis on the rainfall spatio-temporal distribution data of the meteorological monitoring station, establish a rainfall trend prediction model based on the long short-term memory neural network, and evaluate the potential impact degree of rainfall on the target area;

[0056] Based on the geological risk level division result and the potential impact degree of rainfall on the target area, screen the high-risk trigger areas;

[0057] For the high-risk trigger areas, evaluate the landslide occurrence probability by constructing an association model between the groundwater level change rate and the geotechnical strength parameters and using the support vector machine algorithm;

[0058] Use the analytic hierarchy process to analyze the geological risk level division result, the potential impact degree of rainfall on the target area, and the landslide occurrence probability, evaluate the comprehensive risk degree of the target area, and judge whether to trigger a graded early warning signal.

[0059] Specifically, obtaining the time series data of geological deformation in the target area through satellite remote sensing monitoring and using the random forest algorithm to evaluate the feature importance of the geological deformation data to divide the geological risk level of the target area includes:

[0060] Obtain multi-source satellite images and extract the original observation data: Use satellite remote sensing images from multiple sources to cover and monitor the target area to obtain image data of continuous time phases. After the obtained image data is preprocessed (such as geometric correction, atmospheric correction, etc.), it will be used to extract the surface deformation information. During the extraction process, high-precision deformation measurement values are obtained through the differential interferometric synthetic aperture radar technology. Define the following formula to describe the deformation extraction process: ; where represents the difference between the deformation data obtained at two different observation times; represents the observation time The corresponding deformation field data; Indicating the observation time The corresponding deformation field data.

[0061] The obtained difference values Are arranged in chronological order to form time-series observation data at each spatial position, denoted as the original observation data.

[0062] Denoising and interpolation are performed on the original observation data: The extracted original observation data usually contains noise, missing values, and outliers. To construct stable and reliable time-series features, denoising and interpolation processing must be performed on the original observation data. Define the original observation data as ; where And Respectively represent the spatial coordinate indices, representing the row and column positions in the image. The following formula is used for denoising filtering: ; where Represents the original observation data after denoising filtering; Represents the denoising filtering function; Represents the observation time.

[0063] After denoising, for data points with missing time phases, a dual completion method based on spatial and temporal interpolation is adopted. Define the interpolation function as ; where Represents the continuous time-series deformation data after interpolation; Represents the interpolation function that combines spatial neighborhood and time-series information.

[0064] Geological attributes and historical disaster situations are selected and uniformly labeled as features: When constructing a complete feature set, in addition to deformation time-series data, geological attributes and historical disaster records within the target area are also introduced. Geological attributes include formation types, fault distributions, lithologies, etc.; historical disaster situation data involves historical occurrence records of events such as landslides and collapses. To unify various types of data, a comprehensive feature vector is constructed for each monitoring point. Define the comprehensive feature vector of a certain monitoring point as: ; where Represents the comprehensive feature vector of the monitoring point; Represents the deformation rate extracted from the continuous time-series deformation data after interpolation; Represents the local slope value of the target area; Represents the lithology code of the target area; Represents the fault density index within the target area; Represents the historical disaster occurrence frequency.

[0065] After all features are normalized, a unified labeling system is formed to ensure the comparability and effectiveness of each feature in the random forest model.

[0066] Use random forest to evaluate feature importance and screen key factors: Based on the constructed comprehensive feature vector, preliminarily evaluate the geological risks of the target area. Use the random forest algorithm to calculate the importance of each feature to the target variable (i.e., risk level) based on its performance in the decision tree. Use the decrease in the Gini index as the feature importance evaluation index. Define the Gini index before and after the decision tree split as the Gini index of the parent node , the Gini index of the left child node , and the Gini index of the right child node . The corresponding node samples are and . Then the feature importance of a single split can be expressed as:

[0067] ; where represents the importance contribution value of a certain feature in the single decision tree split process; represents the Gini index of the parent node, which is used to measure the impurity of the data before splitting; represents the Gini index of the left child node; represents the Gini index of the right child node; represents the proportion of the sample quantity in the left child node to the total sample quantity of the parent node; represents the proportion of the sample quantity in the right child node to the total sample quantity of the parent node.

[0068] The random forest model will statistically analyze each feature in all decision trees and obtain the overall feature importance score. By setting a predefined threshold, eliminate the features below the predefined threshold, and screen out the key factors that contribute the most to risk assessment. The screening process not only considers the deformation data, but also comprehensively combines geological attributes and historical disaster information, realizing the complementarity and optimization among multi-dimensional features.

[0069] Use a convolutional neural network to identify reflection patterns and interference features: Based on the random forest model, introduce a convolutional neural network to perform deep feature extraction on the deformation image data. Capture the signal reflection pattern and the local features of the interference fringes in the target area through convolution operations. The mathematical expression of the convolution operation is: ; where represents the local features output after the convolution operation; represents the activation function; represents the convolutional kernel, that is, a set of trainable weight matrices; represents the input data matrix, that is, the interpolated continuous time-series deformation data; represents the bias term.

[0070] In a convolutional neural network, through multiple layers of convolution and pooling operations, it is possible to effectively extract reflection patterns and interference features, and fuse these deep features into a comprehensive feature vector, which forms a complement with the previously selected key factors, improving the accuracy and robustness of risk assessment.

[0071] Based on the screening factors, divide the geological risk levels of the target area: construct a risk assessment model, and define the non-linear mapping function as: ; where represents the geological risk level of the monitoring point; represents the interpolated continuous time-series deformation data; represents the comprehensive feature vector of the monitoring point; represents the local features output after convolution operation; represents the non-linear mapping function.

[0072] Mapping function Combines statistical learning and deep feature fusion, outputs a quantitative risk index by weighting the importance of each feature, and then according to the preset risk level division criteria, assigns each location in the monitoring area (i.e., the target area) to different risk levels. The division criteria can be calibrated through experimental data. For example, set the risk thresholds and , when the output of the mapping function is less than the risk threshold , it means that the geological risk level of the target area is level one; when the output of the mapping function is greater than or equal to the risk threshold , and less than the risk threshold , it means that the geological risk level of the target area is level two; when the output of the mapping function is greater than or equal to the risk threshold , it means that the geological risk level of the target area is level three.

[0073] Specifically, perform wavelet transform analysis on the rainfall spatio-temporal distribution data of meteorological monitoring stations, establish a rainfall trend prediction model based on a long short-term memory neural network, and evaluate the potential impact degree of rainfall on the target area, including:

[0074] Collect and integrate the rainfall spatio-temporal distribution data of meteorological monitoring stations in the target area: collect rainfall information through meteorological monitoring stations and sensors in the target area to obtain the rainfall spatio-temporal distribution data of the target area. After the data of each monitoring station is spatio-temporally aligned, it is stored in the form of a digital matrix. Define the rainfall amount data in the form of a function ; where represents the rainfall amount data; represents the spatial coordinate of the monitoring point in the horizontal direction; represents the spatial coordinate of the monitoring point in the vertical direction; represents the rainfall observation time.

[0075] Perform multi-scale time-frequency decomposition on rainfall data using wavelet transform: Considering the non-stationary and multi-scale spatio-temporal characteristics of rainfall data, wavelet transform is used to perform time-frequency decomposition on rainfall data to extract the temporal features at each scale. Introduce the wavelet transform formula:

[0076] ; where represents the wavelet coefficient corresponding to the spatial coordinate , rainfall observation time and scale parameter ; The integration variable, representing continuous time; represents the mother wavelet function, which is used to decompose the input data at different scales; represents the scale parameter, which controls the resolution of the wavelet transform.

[0077] Through the wavelet transform formula, the rainfall data is decomposed into components at multiple scales, providing rich time-frequency information for constructing the prediction model.

[0078] Construct a long short-term memory network model, train and validate the rainfall trend prediction samples: Based on the obtained wavelet decomposition results, use them as the input of the long short-term memory neural network to construct a rainfall trend prediction model. Define the prediction function as ; where represents the rainfall at the spatial coordinate , rainfall observation time ; represents the mapping function of the long short-term memory network model, and its internal structure includes an input gate, a forget gate, and an output gate, which are used to capture the long-range dependencies in the time series; represents the predicted rainfall observation time, which is different from the rainfall observation time , representing the future rainfall observation period predicted by the long short-term memory network model.

[0079] To train the long short-term memory network model, the mean square error is used as the loss function, and its calculation formula is: ; where represents the mean square error value, which is used to measure the gap between the prediction result and the actual observation value; represents the total number of samples, that is, the number of observation points participating in training and validation; represents the actual rainfall of the th sample; represents the predicted rainfall of the th sample by the long short-term memory network model.

[0080] The long short-term memory network model updates its parameters using the backpropagation through time algorithm and continuously reduces the mean squared error value using an adaptive optimization algorithm (such as the Adam optimizer) to achieve a high prediction accuracy. After sufficient training and validation, the model has good rainfall trend prediction ability.

[0081] Calculate the regional rainfall impact coefficient based on the prediction results: Use the rainfall data predicted by the trained long short-term memory network model to evaluate the rainfall impact on the target area. When calculating the rainfall impact coefficient of the target area, set the reference rainfall as the long-term average value and quantify it through the relative change rate. Define the calculation formula for the rainfall impact coefficient as:

[0082] ; where represents the rainfall impact coefficient; represents the th predicted rainfall at the monitoring point; represents the reference value of the rainfall in the target area; represents the total number of effective monitoring points in the target area.

[0083] By statistically analyzing the average relative change of the rainfall deviation from the reference value at each monitoring point, it can intuitively reflect the abnormality degree of rainfall events in the entire target area.

[0084] Evaluate the potential impact degree of rainfall on the target area based on the rainfall impact coefficient: Preset the rainfall impact coefficient threshold and compare the rainfall impact coefficient with the rainfall impact coefficient threshold:

[0085] When the rainfall impact coefficient is greater than the rainfall impact coefficient threshold, it indicates that the potential impact degree of the rainfall event on the target area is high; the rainfall amount and its spatial and temporal distribution in the target area deviate from the historical norm, the rainfall is strong and lasts for a long time, indicating that the soil water absorption capacity in the target area may quickly reach saturation, the surface runoff increases sharply, and the risk of geological disasters rises;

[0086] When the rainfall impact coefficient is less than or equal to the rainfall impact coefficient threshold, it indicates that the potential impact degree of the rainfall event on the target area is low; the rainfall situation in the target area is basically within the historical norm range, and neither the rainfall intensity nor its spatial and temporal distribution reaches the critical state of triggering major disasters, indicating that the rainfall amount in the target area is relatively uniform, the soil moisture and drainage capacity match, the hydrological environment is stable, and the overall risk is low.

[0087] The rainfall impact coefficient threshold is determined based on long-term meteorological observation data, historical disaster records, and regional geological and hydrological characteristics through statistical analysis, regression models, and probability calculations, and a critical value that can accurately distinguish normal and abnormal rainfall states is selected.

[0088] Specifically, based on the results of geological risk level division and the potential impact degree of rainfall on the target area, high-risk trigger areas are screened;

[0089] The results of geological risk level division of the target area include level one, level two, and level three.

[0090] When the geological risk level of the target area is level three, the target area is determined as a high-risk trigger area; when the geological risk level of the target area is level two and the rainfall impact coefficient is greater than the rainfall impact coefficient threshold, the target area is determined as a high-risk trigger area; otherwise, the target area is determined as a non-high-risk trigger area.

[0091] Specifically, for high-risk trigger areas, by constructing a correlation model between the groundwater level change rate and the geotechnical strength parameters, the support vector machine algorithm is used to evaluate the landslide occurrence probability, including:

[0092] Collect groundwater level change rate monitoring data: In the landslide monitoring scenario, the change of groundwater level often affects the pore water pressure of the geotechnical body, thus affecting the stability of the geological body. To accurately obtain the water level change rate at different positions and times, it is necessary to arrange multiple monitoring wells or sensors in the high-risk trigger area to record the groundwater level data for a continuous period. Define the original observation data function as: ; where represents the monitoring value of the groundwater level change rate; represents the well point number, used to distinguish the positions of different observation points; represents the layer number, for example, there are multiple layers of observation points in the same observation well; represents the time index, reflecting the order at different observation times.

[0093] Select geotechnical strength elements and construct correlation factors: In addition to the groundwater level change rate, the geotechnical strength has an important impact on the occurrence of landslides. Geotechnical strength usually includes various mechanical indexes such as the internal friction angle, cohesion, and density. To comprehensively reflect the geotechnical properties, different representative elements are integrated and defined as ; where represents the comprehensive element of the geotechnical strength.

[0094] In engineering practice, if a certain monitoring well corresponds to certain lithological parameters, these parameters are normalized and then uniformly indexed to construct a correlation factor, denoted as .

[0095] Establish a correlation model between the groundwater level and geotechnical strength: There is a non-linear coupling relationship between the groundwater level change rate and the geotechnical strength, and it is necessary to establish a correlation model to obtain a characteristic combination that can represent the landslide risk. Define a correlation function to describe the correlation model between the water level and geotechnical strength:

[0096] ;

[0097] Among them, represents the output value of the water level - geotechnical strength correlation model; represents the correlation function, which is used to perform non - linear mapping on the comprehensive elements of the groundwater level change rate data and the geotechnical strength.

[0098] The output of the correlation model will be used as one of the input features of the support vector machine to characterize the comprehensive state of water level strength at a certain location and at a certain moment.

[0099] The support vector machine is trained by the method of kernel function optimization: In order to evaluate the landslide occurrence probability, a support vector machine classifier is sampled. Since whether a landslide occurs is often in the mode of "occurred / not occurred" or "high - risk / low - risk", a binary classification model can be trained first, and the probability estimation can be derived. Define the training sample set as: ; Among them, represents the th feature vector of the training sample; represents the th class label of the training sample (for example, 1 indicates that a landslide was recorded historically, and 0 indicates no landslide); represents the total number of training samples.

[0100] Combine the output of the correlation model with other auxiliary features (such as vegetation coverage, terrain curvature, etc.) to form a feature vector , and mark the distribution of historical landslide events .

[0101] Train the support vector machine, and its decision function is expressed as:

[0102] ; Among them, represents the output value of the support vector machine decision function for the th training sample (the feature vector is ); represents the Lagrange multiplier corresponding to the th training sample; represents the kernel function, which is used to calculate the similarity between the th training sample and the th training sample in the feature space; represents the bias term, which is the constant term in the support vector machine decision function and is used to adjust the position of the classification boundary.

[0103] The decision function is obtained by accumulating the contributions of all training samples, reflecting the position and distance of the th training sample on the classification plane, providing a basis for landslide risk determination.

[0104] Input the associated model parameters and evaluate the landslide occurrence probability: Probabilize the classification results, and use the Logistic function to map the output value of the support vector machine decision function to the interval (0, 1). Define the probability output as: ; where represents the predicted value of the landslide occurrence probability for the th training sample; represents the exponential function.

[0105] The larger the predicted value of the landslide occurrence probability, the higher the risk of landslide in the high-risk triggering area; at this time, the geological body is prone to instability, the landslide triggering conditions tend to be extreme, and serious geological disasters may be triggered.

[0106] Specifically, use the analytic hierarchy process to analyze the geological risk level classification results, the potential impact degree of rainfall on the target area, and the landslide occurrence probability, evaluate the comprehensive risk degree of the target area, and judge whether to trigger the graded warning signal, including:

[0107] Construct an analytic hierarchy process model, whose top-level goal is to evaluate the comprehensive risk degree of the target area, the middle layer is three risk factors, and the bottom layer is the specific measurement values or quantification results of each risk factor.

[0108] The risk factors include:

[0109] Geological risk level classification results: The target area is divided into first-level, second-level, and third-level, reflecting the inherent stability and vulnerability of the geological body;

[0110] Rainfall influence coefficient: Quantify the potential impact degree of rainfall on the target area;

[0111] Landslide occurrence probability: The predicted value of the landslide occurrence probability calculated by using the support vector machine model, indicating the probability size of the landslide risk.

[0112] In the analytic hierarchy process, it is necessary to make pairwise comparisons of each risk factor to determine the relative importance to the comprehensive risk. The following weights are obtained through expert scoring or statistical analysis:

[0113] Use to represent the weight of the geological risk level classification results; use to represent the weight of the rainfall influence coefficient; use to represent the weight of the predicted value of the landslide occurrence probability. Satisfy .

[0114] Convert the geological risk level classification results into quantitative indicators, defined as: ; where represents the geological risk value of the target area, which is discrete data. For example, 1 represents the first level, 2 represents the second level, and 3 represents the third level.

[0115] Define the comprehensive risk index of the target area as:

[0116] ; where represents the comprehensive risk index; represents the geological risk value of the target area; represents the predicted value of the landslide occurrence probability for the th training sample; represents the weight of the geological risk value; represents the weight of the rainfall influence coefficient; represents the weight of the predicted value of the landslide occurrence probability.

[0117] Preset a comprehensive risk threshold, compare the comprehensive risk index with the comprehensive risk threshold, evaluate the comprehensive risk level of the target area, and determine whether to trigger a graded warning signal:

[0118] When the comprehensive risk index is greater than or equal to the comprehensive risk threshold, it indicates that the comprehensive risk level of the target area is high. At this time, it shows that all risk indicators in the target area have reached or exceeded the set critical level, the target area is in an overall high-risk state, obvious potential safety hazards exist in the target area, it is necessary to trigger a graded warning signal, and take corresponding preventive and emergency measures to effectively reduce the disaster risk;

[0119] When the comprehensive risk index is less than the comprehensive risk threshold, it indicates that the comprehensive risk level of the target area is low, and the geological body is stable under the current environment, and there is no need to trigger a graded warning signal.

[0120] The comprehensive risk threshold is a critical value determined based on long-term historical data, statistical analysis, and expert evaluation, used to delimit the risk boundary of geological disasters occurring in the target area.

[0121] Embodiment 2:

[0122] The difference between Embodiment 2 and Embodiment 1 of the present invention is that this embodiment introduces a geological disaster warning system based on big data analysis.

[0123] Figure 2 The structural schematic diagram of a geological disaster warning system based on big data analysis of the present invention is given. A geological disaster warning system based on big data analysis includes a risk level division module, an impact degree evaluation module, a region screening module, a landslide probability evaluation module, and a warning signal judgment module;

[0124] The risk level division module obtains the time series data of geological deformation in the target area through satellite remote sensing monitoring, and uses the random forest algorithm to evaluate the feature importance of the geological deformation data, and divides the geological risk level of the target area;

[0125] The impact degree evaluation module performs wavelet transform analysis on the rainfall spatio-temporal distribution data of the meteorological monitoring station, establishes a rainfall trend prediction model based on the long short-term memory neural network, and evaluates the potential impact degree of rainfall on the target area;

[0126] The area screening module screens the high-risk trigger areas based on the results of the geological risk level division and the potential impact degree of rainfall on the target area;

[0127] For the high-risk trigger areas, the landslide probability evaluation module evaluates the landslide occurrence probability by constructing an association model between the groundwater level change rate and the geotechnical strength parameters and using the support vector machine algorithm;

[0128] The early warning signal judgment module uses the analytic hierarchy process to analyze the results of the geological risk level division, the potential impact degree of rainfall on the target area, and the landslide occurrence probability, evaluates the comprehensive risk degree of the target area, and judges whether to trigger a hierarchical early warning signal.

[0129] The above formulas are all dimensionless and take their numerical calculations. The formula is a formula obtained by collecting a large amount of data for software simulation to get the closest real situation. The preset parameters and threshold selection in the formula are set by those skilled in the art according to the actual situation.

[0130] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on the computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that the computer can access or a data storage device such as a server or data center that includes one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0131] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0132] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and modules described above can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0133] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division. In actual implementation, there can be other division methods. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the couplings, direct couplings, or communication connections shown or discussed with each other can be through some interfaces. The indirect couplings or communication connections of devices or modules can be in electrical, mechanical, or other forms.

[0134] The modules described as separate components may or may not be physically separated. The components shown as modules may or may not be physical modules. They can be located in one place or distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0135] In addition, the functional modules in each embodiment of this application can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0136] When the above-mentioned functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

[0137] As described above, the above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed by this application can easily think of changes or substitutions, which should all be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0138] Finally: The above is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A geological disaster warning method based on big data analysis, characterized in that, It includes the following steps: Obtain the time series data of geological deformation in the target area through satellite remote sensing monitoring, and use the random forest algorithm to evaluate the feature importance of the geological deformation data to divide the geological risk level of the target area; Conduct wavelet transform analysis on the rainfall spatio-temporal distribution data of the meteorological monitoring station, establish a rainfall trend prediction model based on the long short-term memory neural network, and evaluate the potential impact degree of rainfall on the target area; Collect and integrate the rainfall spatio-temporal distribution data of the meteorological monitoring station in the target area; Perform multi-scale time-frequency decomposition on the rainfall amount data using wavelet transform; Construct a long short-term memory network model, train and verify the rainfall trend prediction samples; Calculate the regional rainfall impact coefficient according to the prediction results; Evaluating the potential impact degree of rainfall on the target area based on the rainfall impact coefficient: Define the calculation formula of the rainfall impact coefficient as: ; where represents the rainfall impact coefficient; represents the predicted rainfall at the th monitoring point; represents the baseline value of rainfall in the target area; represents the total number of effective monitoring points in the target area; Preset the rainfall impact coefficient threshold, and compare the rainfall impact coefficient with the rainfall impact coefficient threshold: When the rainfall impact coefficient is greater than the rainfall impact coefficient threshold, it indicates that the potential impact degree of the rainfall event on the target area is high; When the rainfall impact coefficient is less than or equal to the rainfall impact coefficient threshold, it indicates that the potential impact degree of the rainfall event on the target area is low; Based on the geological risk level division result and the potential impact degree of rainfall on the target area, screen the high-risk trigger areas; For the high-risk trigger areas, evaluate the landslide occurrence probability by constructing an association model between the groundwater level change rate and the rock and soil mass strength parameters and using the support vector machine algorithm; Collect the groundwater level change rate monitoring data to obtain the groundwater level change rate monitoring value; Select the rock and soil mass strength elements and construct the association factors; Establish a groundwater level-rock and soil strength association model; define an association function to describe the groundwater level-rock and soil strength association model; the association factors and the groundwater level change rate monitoring value will be used as the input of the association model; Train the support vector machine by using the kernel function optimization method; the output of the association model will be used as the input feature of the support vector machine; Input the association model parameters to evaluate the landslide occurrence probability; the association model parameters are the output values of the support vector machine decision function, and the Logistic function is used to map the output values of the support vector machine decision function to the interval (0,1); Use the analytic hierarchy process to analyze the geological risk level division result, the potential impact degree of rainfall on the target area, and the landslide occurrence probability, evaluate the comprehensive risk degree of the target area, and judge whether to trigger a graded warning signal.

2. The geological disaster early warning method based on big data analysis according to claim 1, wherein, Obtain the time series data of geological deformation in the target area through satellite remote sensing monitoring, and use the random forest algorithm to evaluate the feature importance of the geological deformation data to divide the geological risk level of the target area. Specifically: Obtain multi-source satellite images and extract the original observation data; Denoise and interpolate the original observation data to obtain continuous time series deformation data; Select the geological attributes and historical disaster situations, uniformly label the features to obtain a comprehensive feature vector; Use the random forest to evaluate the feature importance and screen the key factors; based on the constructed comprehensive feature vector, use the random forest algorithm; by setting a predefined threshold, eliminate the features below the predefined threshold, and screen out the key factors that contribute the most to the risk assessment; Identifying reflection patterns and interference features using a convolutional neural network; the input of the convolutional neural network is continuous time-series deformation data, and the output is local features; the reflection patterns and interference features refer to the local features of the signal reflection pattern and interference fringes within the target area. Based on screening factors, classifying the geological risk levels of the target area; the screening factors include local features and key factors that contribute the most to risk assessment selected from the comprehensive feature vectors.

3. A geological disaster early warning method based on big data analysis according to claim 2, characterized in that, Based on the classification result of the geological risk level and the potential impact degree of rainfall on the target area, screening high-risk trigger areas, specifically: When the geological risk level of the target area is level three, determining that the target area is a high-risk trigger area; When the geological risk level of the target area is level two and the rainfall impact coefficient is greater than the rainfall impact coefficient threshold, determining that the target area is a high-risk trigger area; Otherwise, determining that the target area is a non-high-risk trigger area.

4. The geological disaster warning method based on big data analysis according to claim 3, wherein, Using the analytic hierarchy process to analyze the classification result of the geological risk level, the potential impact degree of rainfall on the target area, and the landslide occurrence probability, evaluating the comprehensive risk degree of the target area, and determining whether to trigger a hierarchical warning signal, specifically: Presetting a comprehensive risk threshold, comparing the comprehensive risk index with the comprehensive risk threshold: When the comprehensive risk index is greater than or equal to the comprehensive risk threshold, it indicates that the comprehensive risk degree of the target area is high and a hierarchical warning signal needs to be triggered; When the comprehensive risk index is less than the comprehensive risk threshold, it indicates that the comprehensive risk degree of the target area is low and a hierarchical warning signal does not need to be triggered.

5. A geological disaster early warning system based on big data analysis, which is used to implement a geological disaster early warning method based on big data analysis according to any one of claims 1-4, and is characterized in that, Including a risk level classification module, an impact degree evaluation module, a region screening module, a landslide probability evaluation module, and a warning signal judgment module; The risk level classification module obtains the geological deformation time-series data of the target area through satellite remote sensing monitoring, uses the random forest algorithm to evaluate the feature importance of the geological deformation data, and classifies the geological risk level of the target area; The impact degree evaluation module performs wavelet transform analysis on the rainfall spatio-temporal distribution data of the meteorological monitoring station, establishes a rainfall trend prediction model based on the long short-term memory neural network, and evaluates the potential impact degree of rainfall on the target area; The region screening module screens high-risk trigger areas based on the classification result of the geological risk level and the potential impact degree of rainfall on the target area; For high-risk trigger areas, the landslide probability evaluation module evaluates the landslide occurrence probability by constructing an association model between the groundwater level change rate and the geotechnical strength parameters and using the support vector machine algorithm; The warning signal judgment module uses the analytic hierarchy process to analyze the classification result of the geological risk level, the potential impact degree of rainfall on the target area, and the landslide occurrence probability, evaluates the comprehensive risk degree of the target area, and determines whether to trigger a hierarchical warning signal.

Images